For more than two decades researchers have utilized the snowmelt runoff model (SRM) to test the impacts of climate change on streamflow of snow-fed systems. SRM developers recommend a parameter shift during simulations of future climate, but this is often omitted. Here we show the impact of this omission on model results. In this study, the hydrological effects of climate change are modeled over three sequential years with typical and recommended SRM methodology. We predict the impacts of climate change on water resources of five subbasins of an arid region. Climate data are downscaled to weather stations. Period change analysis gives temperature and precipitation changes for 55 general circulation models which are then subsampled to produce four future states per basin. Results indicate an increase in temperature between 3.0 and 6.2 °C and an 18% decrease to 26% increase in precipitation. Without modifications to the snow runoff coefficient (cS), mean results across all basins range from a reduction in total volume of 21% to an increase of 4%. Modifications to cS resulted in a 0–10% difference in simulated annual volume. Future application of SRM should include a parameter shift representing the changed climate.

INTRODUCTION

The impact of climate change on water resources of presently water-limited regions represents a critical research topic. The quantity of freshwater available per person is decreasing due to a combination of factors including population growth, water pollution and inadequate water management and operations (Sivakumar 2011). Global climate change will cause more frequent and severe hydrological extremes including floods and droughts, compounding water management challenges (Sivakumar 2011; Garfin et al. 2013). In the southwestern United States, scientists report with high confidence that warming will continue and summer heat waves will be longer and hotter. As a consequence, late-season snowpack, which supplies water to much of the region, will decrease. Decline of water availability in limited regions impacts ecosystems and the humans relying upon those ecosystems. In some arid regions, a decline in available freshwater could spur major transitions as those relying on water to grow crops, support livestock and supply human uses continue to cope with larger demand than supply. In many regions of the world future climate is predicted to be hotter and drier, thereby further limiting critical water resources (Kudzewiczn et al. 2008). Future warmer temperatures in the arid southwestern United States will alter the hydrologic cycle by increasing potential evapotranspiration and causing earlier snowmelt, thus shifting timing of available water.

The main water source for many arid regions, such as the Rio Grande in the southwestern United States, is upstream snowpack (Rango 2006). Snow covered areas (SCAs) are particularly vulnerable to increased temperatures of a future climate. The snowmelt runoff model (SRM) is designed to simulate the impacts of temperature and precipitation change on snowpack, streamflow and water volume. In fact, the first simulation using estimated future temperatures to predict streamflow with SRM occurred in 1980 (Martinec 1980). Since then, many institutions worldwide have made major advances in future climate simulation with atmospheric-oceanic general circulation models (GCMs) and coordinated efforts through the multi-model intercomparison projects (Coupled Model Intercomparison Project (CMIP)). With the CMIP3 and CMIP5 models, we now have many historic and future projections to inform hydrologic research. Here we utilize a methodology to statistically subsample available models (Mejia et al. 2011) and simulate the impacts of climate change on hydrology with SRM in a highly vulnerable region (Overpeck et al. 2013).

Recent observations in the southwestern United States indicate that the region is already warming, experiencing unusually severe droughts and diminished streamflows (Overpeck et al. 2013). Analysis of historic regional temperature trends during the last three decades show mean annual temperatures increases of about 0.5–1 °C per decade (Rangwala & Miller 2012). A recent climate risk assessment in this area anticipates temperature increases of 4–6 °C by the end of the 21st century, and this will have deleterious effects on streamflow and water supplies (Llewellyn & Vaddey 2013). Increased warming and decreased snowpack forewarn a likely continued decline in the available water of the Rio Grande.

The SRM, a temperature index model designed to simulate snowmelt in mountainous areas, is well suited to utilize future projections to estimate the impacts of climate change on hydrology. For more than two decades researchers have employed SRM in basins around the globe to predict the hydrologic and associated effects of warming temperatures and changes in precipitation. Although model developers recommend a parameter shift when simulating climate change (Martinec et al. 2008), this shift is rarely utilized, possibly leading to erroneous estimates of future temperatures on hydrology.

This research aims to:

  1. apply a subsampling methodology to statistically select future climate models for SRM simulation;

  2. utilize SRM with four future conditions to predict the range of impacts of climate change on SCA and hydrology with the typical application of SRM (no parameter shift);

  3. employ recommended methodology (parameter shift) in climate change simulations to estimate the range of future climate impacts on SCA and hydrology.

Research to reduce uncertainty and provide for a more reliable assessment of how climate change may affect water resources is urgently needed (Kudzewiczn et al. 2008). The methodology presented here, both subsampling all climate models and use of recommended SRM methodology, address this urgent research need.

SRM

SRM is a temperature index model developed by Martinec (1975) to simulate daily streamflow in basins where snowmelt contributes much of the annual runoff (Martinec 1975). Over time SRM has been widely used to simulate the effects of a variety of hydrological applications including streamflow estimation, flood forecasting, optimizing reservoir operations, improved power generation and climate change simulations. The model has been applied in over 100 basins in at least 29 different countries (Martinec et al. 2008). SRM was initially applied to small, mountainous European basins. Researchers then simulated larger basins with SRM and to date it has been used in basins ranging from 0.8 to 917,444 km2. SRM successfully underwent tests by the World Meteorological Organization (WMO) with regard to runoff simulations (WMO 1986) and streamflow forecasting (WMO 1992). As knowledge of climate change has expanded over the past decades, SRM has proven useful in evaluating the hydrologic impacts of warming temperatures and changing precipitation patterns.

Researchers often use SRM to evaluate the effects of climate change (Martinec 1980; Rango & van Katwijk 1990, 1991; Nash & Gleick 1991; van Katwijk & Rango 1991; van Katwijk et al. 1993; Gyalistras et al. 1994; Rango & Martinec 1994, 1999, 2000; Seidel et al. 2000; Wang et al. 2010; Tahir et al. 2011; Sharma et al. 2013; Elias et al. 2015). Following initial runoff simulation to verify model efficiency, climate change simulations in SRM are relatively simple. It is possible to select changes in temperature or precipitation and apply those changes for a designated period of time. Results show the summer and winter effects of a changed climate on SCA and runoff.

In a series of articles focused on simulating the hydrologic effects of climate change with SRM, Rango & van Katwijk (1990) first simulate climate change with no change in parameter timing. They found that the major effect is a result of increased temperatures, shifting the runoff to earlier in the year and increased streamflow in April and May. Subsequent research employs stepwise changes in model parameters wherein they found that an increase in temperature will cause a time shift in certain model parameters that somewhat reduces the previously reported increase in April and May flows while magnifying the decreased summer streamflow (Rango & van Katwijk 1991). Since temperature increase causes an earlier start to snowmelt, the regular sequence of basin conditions related to losses, ripeness of snowpack and snow density will advance accordingly. The impact of shifting six SRM parameters (snow runoff coefficient, rain runoff coefficient, degree-day factor, rainfall contribution area, lapse rate and critical temperature) was then tested on two basins (van Katwijk & Rango 1991). The snow runoff coefficient had by far the largest effect on altering seasonal runoff volume. Authors conclude that ‘there is a significant effect of time shifting the model parameters to coincide with the climate change conditions'. The average effect of parameter shifts was a 5–10% decrease in total runoff volume. A fourth article in the series concludes that it is ‘very important to include the expected change in climate related basin conditions' (van Katwijk et al. 1993). The impact of shifting model parameters to represent basin conditions caused higher streamflow in April and even lower June and July total volume results. The practical implications of omitting this parameter shift include underestimation of climate change related flooding events and overestimation of available future water resources in the months of June and July, when irrigation demand and evaporative water losses are already high. Runoff overestimation by failing to shift parameters while evaluating the impacts of increased temperatures could lead to critical errors, especially in arid regions where water demand presently exceeds supply.

Due to climate change, snowmelt runoff research is essential for predicting water availability. In arid lands where mountain fed rivers are the only available surface water resources to supply the public, developing best practices to provide water on a sustainable and long-term basis appears more urgent (Abudu et al. 2012). The Upper Rio Grande (URG) is one such region where snow-fed rivers provide for downstream inhabitants and agricultural production. Accurate simulation of snowmelt runoff is a necessity for this region.

Although model developers recommend a parameter shift to best represent conditions in a changed climate, this procedure is not applied in more recent applications of SRM used to assess climate change impacts (Wang et al. 2010; Tahir et al. 2011; Sharma et al. 2013). Recent SRM simulation of the URG did utilize a parameter shift on 24 simulated basins (Elias et al. 2015). Here we simulate the hydrologic impact of climate change with and without recommended parameter shifts to illustrate the effect of holding parameters constant during climate change simulations, while making use of recent future climate projections not available during prior SRM-climate change simulations, to define best practices for future applications of SRM in climate change simulation.

Study region: Rio Grande Basin

The URG Basin is located in the semi-arid southwestern United States. Snowfall in the San Juan and Sangre de Cristo Mountains of southern Colorado and northern New Mexico drains to the mainstem of the Rio Grande. This research focuses on five basins located in the mountainous headwaters of the San Juan Range (Figure 1). We selected the five most productive subbasins in the URG, generally providing more than 70% of the annual runoff volume. The most important source of water in the Rio Grande drainage results from snowmelt in the mountains of the upper basin as 50–75% of the flow in the Rio Grande is sustained by melting snow (Rango 2006).
Figure 1

URG study region located in the San Juan Mountains of Southern Colorado and Northern New Mexico, USA.

Figure 1

URG study region located in the San Juan Mountains of Southern Colorado and Northern New Mexico, USA.

The snowmelt dominated Rio Grande streamflow generally peaks in the late spring and early summer and diminishes rapidly by mid-summer. However, the highest evapotranspiration and irrigation demands along the Rio Grande occur from June through mid-September, when flows are characteristically low (Llewellyn & Vaddey 2013). Streamflow in the URG is historically highly variable and droughts, defined as a year or more with annual flows less than the long-term median, are common (Woodhouse et al. 2012).

By the end of the century, temperatures in the URG are anticipated to increase by about 5 °C under high emissions global climate model scenarios (Cayan et al. 2013). Temperature increases will be greatest in summer and fall. While models are split between those showing declines in winter precipitation and those showing small increases, winter precipitation is expected to increasingly fall as rain rather than snow (Gutzler et al. 2006). Temperature driven increases in evaporation will change the components of the overall water budget, resulting in less available water even with potential small increases in precipitation (Nash & Gleick 1993). Given the large percentage of Rio Grande streamflow derived from snowmelt, simulation of a snowmelt and streamflow response to anticipated increased temperatures of a changed climate is vital for developing adaptive management strategies. Water resources of this region are particularly vulnerable to the projected increased temperatures since supplies are presently limited. Increased temperatures and population growth in the Rio Grande basin will cause the gap between water supply and demand to continue to grow (Rango 2006). The timing of water supply will shift to earlier in the year and water management flexibility for current water users will decrease (Elias et al. 2015). Additionally, the notion that groundwater supplies can be tapped to make up the deficit in future shortages ignores groundwater supply limitations as the groundwater reservoir is already heavily mined and depleted in the basin (Bartolino & Cole 2002).

METHODS

This section describes the methods used to downscale GCM data, simulate base and climate change scenarios with SRM and shift the runoff coefficient (cS).

Downscaled GCM data

GCM temperature and precipitation data were obtained from the statistically downscaled products known as Bias-Correction Constructed Analogues BCCA-CMIP3 (Maurer et al. 2010) and BCCA-CMIP5 (Reclamation 2013) at daily time increments and grid size of 1/8° (∼10 km). However, these products are too coarse in space to supply useful information for SRM. As implemented in this paper, calibration and simulation with SRM used direct meteorological point measurements. Over complex terrain the mean historical BCCA products are still biased relative point measurements due to insufficient spatial resolution. Previous studies showed different remaining challenges associated with application of downscaling approaches (∼10 km) for hydrological modeling on small-scale subbasins, especially for complex terrain (Mejia et al. 2011; Huntington & Niswonger 2012). Hence, for consistency and to further improve BCCA products, it was necessary to post-process and project the data to the same point measurement locations. For this, the station-based probability distribution was implemented to perform a quantil-quantil distribution mapping approach as described in Mejia et al. (2014). The combination of BCCA products and this additional point measurement downscaling approach is termed here as ‘double-statistical downscaling’. This post-processing is further justified in regions where data is sparse, and large spatial gradients of meteorological fields exist, as is the case in the URG Basin, limiting the use of gridded products (Gudmundsson et al. 2012; McEvoy et al. 2014; Haerter et al. 2015). Of note is that hydrologic simulations using similar double-statistical downscaling approach quantitatively improve historical hydrologic simulations for most evaluation indicators, and yield unbiased results of hydrologic projections in small catchments (Huntington & Niswonger 2012; Mejia et al. 2014). Since SRM base simulations use measured data from specific weather stations, capturing future changes at those stations is necessary to assess the impacts of climate change.

Subsampling GCM

Often climate assessment studies use composites or subsamples of CMIP3 and CMIP5 according to their individual skill evaluated by region or by individual or combination of parameters (Knutti et al. 2010). Ensemble subsample selection, or weighting depending ensemble member performance, is highly sensitive of the intended application, region of interest, and the metrics implemented (Perkins et al. 2012). Even though the skill of model results and applications can improve by using such ensemble or composited approaches (Stott & Forest 2007; Knutti et al. 2010), these improvements may change under different time slices, the unlimited parameter/metric options for GCM skill evaluation, which also requires accurate knowledge of the single model skill (Weigel et al. 2010). Additionally, GCMs suffer from systematic biases, which indicate that models have physics limitations. For example, precipitation over the western USA tends to be over-predicted in most GCM (Dai 2006; Sheffield et al. 2013). Further, skill scores based on downscaled approaches (e.g. BCCA, double-statistical approach), which tend to underscore the inherited limitations from their parent GCMs, would preclude due to the underlying assumptions in the mapping procedures that forcefully remove the original biases.

Limitations in computing resources require an efficient and unbiased subsample strategy, which also allows characterizing the GCMs/downscaled inherited range of uncertainty in the weather data. For these purposes, we adopted a similar approach to the method proposed by Reclamation (2008), consisting of selection of the four intersecting points defined by the nearest GCMs intersection of 10th to 90th percentile changes in temperature and 10th to 90th percentile changes in precipitation and combining all the scenarios available. We selected the 10th and 90th percentile changes of the temperature-precipitation parameter space representing warmer/wetter (WW), warmer/drier (WD), hotter/wetter (HW), and hotter/drier (HD) mean future projections. Of note is that this approach is robust and covers the envelope of weather possibilities, minimizes information loss, and assumes that any combination of precipitation-temperature states within this envelope creates a hydrologic response that falls within the range of hydrologic solutions produced by the four extreme ensemble members. The major difference between our approach and that of Reclamation (2008) is that we opted for the warmest climate scenario (CMIP3-A2 and CMIP5-RCP8.5).

SRM methodology

SRM was used to simulate measured streamflow at the outlet of five subbasins tributary to the URG using observed precipitation and temperature data collected at climate stations within or near the subbasins. Necessary data, SRM set-up and climate simulation methods are described below. A detailed description of SRM methodology and data requirements is presented elsewhere (DeWalle & Rango 2008; Martinec et al. 2008).

Basin characteristics

Basin characteristics required for SRM include basin area, gage location at the basin outlet, gage elevation, elevation zones, zone areas, and basin or zonal hypsometric mean elevation. We acquired gage locations and elevations from the US Geological Survey (USGS) National Water Information Service (NWIS) website (USGS 2001) and delimited the study basins above stream gage locations using National Elevation Dataset (NED) digital elevation data (Gesch et al. 2002; Gesch 2007) and ArcGIS Hydrology tools (ESRI 2010). We also used digital elevation data to divide basins into elevation zones (∼500 m elevation range in each zone) and to calculate area and hypsometric mean elevations of each elevation zone (Table 1). Either 1-arc second (30 m) NED data (Gesch et al. 2002; Gesch 2007) or 1-arc second Shuttle Radar Topography Mission data (USGS 2012) were used for this research.

Table 1

Subbasin gage names and elevations, elevation zones, area and weather stations

Gage name, USGS site number Elevation (m.a.s.l.) Area (km2Weather stations 
Rio Grande Del Norte: 08220000 2,432 3,416 Del Nortea, Wolf Creekc Summitc 
Zone A 2,436–2,926 779 
Zone B 2,927–3,353 1,282 
Zone C 3,354–4,222 1,354 
Rio Chama near La Puente: 08284100 2,159 1,222 Chamitac, Cumbres Trestlec 
Zone A 2,159–2,750 487 
Zone B 2,750–3,325 692 
Zone C 3,325–3,886 50 
Conejos River near Mogote: 08246500 2,522 729 Big Hornb, Lily Pondc 
Zone A 2524–2925 146 
Zone B 2,925–3,350 323 
Zone C 3,350–4,005 259 
Los Piños River near Ortiz: 08248000 2,451 395 Big Hornb, Cumbres Trestlec 
Zone A 2,454–2,900 115 
Zone B 2,900–3,250 234 
Zone C 3,250–3,716 46 
Alamosa River above Terrace Reservoir: 08236000 2,621 274 Del Nortea, Lily Pondc 
Zone A 2,624–2,925 25 
Zone B 2,925–3,350 106 
Zone C 3,350–4,036 143 
Gage name, USGS site number Elevation (m.a.s.l.) Area (km2Weather stations 
Rio Grande Del Norte: 08220000 2,432 3,416 Del Nortea, Wolf Creekc Summitc 
Zone A 2,436–2,926 779 
Zone B 2,927–3,353 1,282 
Zone C 3,354–4,222 1,354 
Rio Chama near La Puente: 08284100 2,159 1,222 Chamitac, Cumbres Trestlec 
Zone A 2,159–2,750 487 
Zone B 2,750–3,325 692 
Zone C 3,325–3,886 50 
Conejos River near Mogote: 08246500 2,522 729 Big Hornb, Lily Pondc 
Zone A 2524–2925 146 
Zone B 2,925–3,350 323 
Zone C 3,350–4,005 259 
Los Piños River near Ortiz: 08248000 2,451 395 Big Hornb, Cumbres Trestlec 
Zone A 2,454–2,900 115 
Zone B 2,900–3,250 234 
Zone C 3,250–3,716 46 
Alamosa River above Terrace Reservoir: 08236000 2,621 274 Del Nortea, Lily Pondc 
Zone A 2,624–2,925 25 
Zone B 2,925–3,350 106 
Zone C 3,350–4,036 143 

aCOOP station.

bRAWS station.

 cDenotes SNOTEL station.

Streamflow data

Streamflow data for Rio Chama and Los Pinos were downloaded from the USGS NWIS website (USGS 2001). Streamflow data for subbasins in Colorado (Del Norte, Alamosa and Conejos) were downloaded from the Colorado Division of Water Resources (2008). Apart from initial flow, streamflow data are not used directly in simulations; rather, these data are used to judge the success of the simulation in computation of model efficiency statistics.

Meteorological data

We used three sources for the daily time-step meteorological data: the National Weather Service Cooperative Observer Program (COOP) (NOAA 2013), the Natural Resources Conservation Service Snow Telemetry (SNOTEL) system (USDA NCRS 2013) and US Forest Service weather stations that are operated as part of the interagency Remote Automatic Weather Station (RAWS) network (USFS 2011). For each basin, we use temperature and precipitation data from two weather stations within or near the subbasin (Table 1). There are relatively few weather stations in the URG. Station elevations are averaged to calculate the elevation of a ‘synthetic’ station. SRM lapses temperatures calculated for the synthetic station to the hypsometric mean of each elevation zone at a rate of 0.6 to 0.8 per 1,000 m. Mean precipitation at the synthetic station is adjusted to the hypsometric mean elevation of each zone by 3.5–4% per 100 m difference in elevation (Martinec et al. 2008).

SCA

Errors in determining SCA are directly proportional to the resulting errors in calculated snowmelt (Rango & Martinec 1981). We previously evaluated snow cover estimates from two remote sensing data sources: the routinely available Terra MODIS 500 m fractional and binary daily snow cover products from the National Snow and Ice Data Center (Hall et al. 2006) and binary maps created from in-house supervised classification of Landsat Thematic Mapper (TM) imagery (Steele et al. 2015, under review). The in-house classification of Landsat TM imagery is more accurate, so we use it to generate estimates of SCA for each zone in each subbasin. Daily estimates of the percent of the zone covered by snow (conventional depletion curves) are then used in SRM simulations.

Model parameters

The model parameters required by SRM include the temperature lapse rate, degree day factor, critical temperature and snow and rainfall runoff coefficients. In SRM all parameters may be varied over the year and by zone to better replicate how hydrological conditions vary throughout the year. Ideally, SRM parameter values should be derived from field measurements or hydrological judgments and not optimized through calibration with streamflow (Martinec et al. 2008). This is because parameter calibration may yield unrealistic values, especially if the model is adjusted to capture extreme events. At the same time, calibration is necessarily a part of hydrologic modeling to achieve the best fit between measured and simulated streamflow while maintaining realistic parameter values. Some studies have used automatic algorithms for parameter calibration of SRM (Boudhar et al. 2010; Panday et al. 2013).

In this study we derived parameter starting values for one year using the methods based on measured data (Martinec et al. 2008). Where the combination of parameters appeared to under- or over-predict streamflow, we adjust parameters, such as rain and snow runoff coefficients, within realistic ranges. SRM parameter adjustment occurs on a basin-by-basin manner and adjustments are made until measured and simulated runoff attains an acceptable agreement for the simulated year. Although the calibration and validation process is not generally utilized with SRM (Martinec et al. 2008), here we undergo additional simulations for verification. We simulate several additional years with the same parameters to expand the number of years upon which we can conduct climate change simulations. In prior work, snow depletion curves were generated for multiple years for each basin to represent a variety of streamflow conditions. In simulating additional years for each basin, parameters were not adjusted, but another year's snow depletion curve, corresponding to the year's streamflow conditions, was occasionally substituted to improve model performance prior to climate change simulation. Depletion curves from a year with similar flow conditions were substituted to improve climate change simulations for Alamosa (WY98) and Del Norte (WY98). Both basins were calibrated to a year of higher flow than that of WY98 and substituting a depletion curve from a lower water year, though computed using the same methodology outlined above, improved simulation to satisfactory model efficiency prior to climate change modeling. As is recommended in hydrologic modeling, percent volume difference and Nash–Sutcliffe Efficiency (NSE) are calculated to verify model efficiency (Moriasi et al. 2007). NSE (Ef) is calculated using daily data and represents the distribution of predicted versus observed discharge (Nash & Sutcliffe 1970; McCuen et al. 2006). Ef values range between 0 and 1 with higher values indicating better agreement between predicted and observed values.

Climate change simulation

We simulate climate change for three years on five URG subbasins. Following basin set-up and parameterization, WinSRM allows for climate change scenario definition within the model. First, the predicted increase in temperature and change in precipitation for each basin and year is defined in the ‘climate scenario definition’ window. In this study, four scenarios are defined for each basin representing the WW, WD, HW and HD future condition. Next, climate change simulations are conducted for each basin, scenario and year in a step-wise manner for the water year with WinSRM (Martinec et al. 2008). Model results and cumulative depletion curves (CDCs) are then evaluated to estimate the shift in runoff timing. Since the initial CDCs are shifted about one month earlier, the values of the runoff coefficient for snow (cS) are shifted accordingly in the climate run. A second climate change simulation is conducted to include a shift in the runoff coefficient for snowmelt 30 days earlier in the water year. The snow runoff coefficient was selected as the most important parameter to shift because it was identified to produce the most effect on runoff in prior research (Rango & van Katwijk 1991). CS reflects both the decline of the snow coverage and the stage of vegetation growth. CS is a coefficient used in SRM to account for all the water losses to represent the difference between the available water (snowmelt and rainfall) and the basin outflow. CS accounts for the losses of sublimation, evaporation, transpiration and groundwater infiltration. Understandably, this coefficient represents multiple physical processes within the watershed and changes spatially (within each zone) and temporally (over the year). In SRM, a different cS value can be used for each day and zone to better represent changes in physical processes. For example, at the start of the snowmelt season, water losses from the basin are small and related mostly to snow surface evaporation and the corresponding higher cS value (∼0.8). The physical changes of warmer temperatures, longer days and proceeding snowmelt also lead to basin plant growth and increased evaporation, evapotranspiration and interception, all processes widening the gap between subbasin input (rain and snow) and outflow. Consequently, to represent these increased water losses within the subbasin, cS values decrease.

RESULTS AND DISCUSSION

SRM streamflow simulation without climate change

Model performance using constant parameters for alternate years is reported in Table 2. For the calibration years, the percent difference between measured and simulated volume was less than 5% and the NSE was between 0.93 and 0.95, indicating very good model performance. The 1997–1998 simulations for Del Norte and Alamosa were improved by using a depletion curve generated for a year with lower streamflow than the calibration year. While model performance is generally good, especially because no parameter adjustment occurs during verification years, Rio Chama simulations generally under-predicted streamflow and have a lower NSE than simulations for other basins (0.72). The Rio Chama models under-predict peak snowmelt in all three verification years, possibly related to the 2000–2001 snow cover. Since verification simulations use 2001 snow cover, if it were too low for one or more zones, then corresponding peak snowmelt runoff would be low. Models used in climate change simulations have a percent difference less than 18% and an Ef greater than 0.80.

Table 2

Evaluation of SRM application during the hydrological year (October–September) for basins of the URG

      Model efficiency
 
  
  Measured volume Computed volume Difference in volume (%) Nash–Sutcliffe coefficient CDC used 
Basin 106 m3 106 m3 – – – 
Alamosa 
 Calibration (98–99) 102 98 0.95 98–99a 
 Verification (96–97) 113 131 –16 0.88 98–99a 
 Verification (97–98) 82 101 –23 0.71 98–99 
 Verification (97–98) 82 79 0.83 88–89a 
Chama 
 Calibration (00–01) 278 291 –5 0.94 00–01a 
 Verification (96–97) 392 273 30 0.71 00–01a 
 Verification (97–98) 320 227 29 0.73 00–01a 
 Verification (98–99) 312 239 23 0.72 00–01a 
Conejos 
 Calibration (98–99) 278 271 0.95 98–99a 
 Verification (96–97) 334 322 0.94 98–99a 
 Verification (97–98) 250 262 –5 0.76 98–99a 
Los Pinos 
 Calibration (98–99) 97 100 –3 0.93 98–99a 
 Verification (97–98) 125 108 14 0.86 98–99a 
 Verification (98–99) 94 91 0.92 98–99a 
Rio Grande at Del Norte 
 Calibration (86–87) 1,324 1,269 0.95 86–87a 
 Verification (96–97) 1,075 1,279 –18 0.82 86–87a 
 Verification (97–98) 786 1,072 –36 0.39 86–87 
 Verification (97–98) 786 677 14 0.80 88–89a 
 Verification (98–99) 1,124 1,303 –15 0.93 86–87a 
      Model efficiency
 
  
  Measured volume Computed volume Difference in volume (%) Nash–Sutcliffe coefficient CDC used 
Basin 106 m3 106 m3 – – – 
Alamosa 
 Calibration (98–99) 102 98 0.95 98–99a 
 Verification (96–97) 113 131 –16 0.88 98–99a 
 Verification (97–98) 82 101 –23 0.71 98–99 
 Verification (97–98) 82 79 0.83 88–89a 
Chama 
 Calibration (00–01) 278 291 –5 0.94 00–01a 
 Verification (96–97) 392 273 30 0.71 00–01a 
 Verification (97–98) 320 227 29 0.73 00–01a 
 Verification (98–99) 312 239 23 0.72 00–01a 
Conejos 
 Calibration (98–99) 278 271 0.95 98–99a 
 Verification (96–97) 334 322 0.94 98–99a 
 Verification (97–98) 250 262 –5 0.76 98–99a 
Los Pinos 
 Calibration (98–99) 97 100 –3 0.93 98–99a 
 Verification (97–98) 125 108 14 0.86 98–99a 
 Verification (98–99) 94 91 0.92 98–99a 
Rio Grande at Del Norte 
 Calibration (86–87) 1,324 1,269 0.95 86–87a 
 Verification (96–97) 1,075 1,279 –18 0.82 86–87a 
 Verification (97–98) 786 1,072 –36 0.39 86–87 
 Verification (97–98) 786 677 14 0.80 88–89a 
 Verification (98–99) 1,124 1,303 –15 0.93 86–87a 

aUsed in climate change simulations.

Climate change scenarios

Four future conditions are determined by subsampling the GCM to find the points representing the bounds of future temperature and precipitation (Figure 2). Data downscaled to six climate stations are used to conduct period change analysis. The models were statistically selected to represent the four scenarios for each weather station (Table 3). The mean 1990–1999 vs. 2090–2099 temperature and precipitation are computed for each model and weather station used in initial simulations for the five URG subbasins. Past simulations have estimated a temperature increase by the end of the 21st century (generally 4 °C) (Rango & Martinec 1997). Estimating future temperatures using data downscaled to the weather station better captures the influence of local physiographic factors to produce more accurate results than using one temperature increase across all stations. Results at the end of the 21st century based upon the RCP8.5 and A2 projections will be more aggressive than mid-century projections with lower greenhouse gas concentration assumptions. The results presented here assume emissions continue to rise throughout the 21st century. Temperature changes to represent the future climate increase ranging from 3.0 to 6.3 °C warmer (Table 3). In contrast, future precipitation indicates both wetter and drier conditions ranging from an 18% decrease in precipitation to a 27% increase in precipitation. The statistically selected models representing each future state for each weather station are reported in Table 4. The HW and HD future conditions for each weather station are represented by models from phase 5 of the CMIP. The WW scenario for each weather station is represented by the same model from phase 3 of the CMIP.
Table 3

Statistically selected models from the CMIP phase 3 A2 scenario and the phase 5 RCP8.5 scenario for each weather station used in SRM simulation

Weather station HD HW WW WD 
Del Norte COOP CSIRO-Mk3-6-0 Run 6 CanESM2 Run 5 MRI-CGCM2.3.2aRun 5 GFDL-ESM2M 
Wolf Creek SNOTEL IPSL-CM5A-LR Run 2 CSIRO-Mk3-6-0 Run 5 MRI-CGCM2.3.2aRun 5 CGCm3.1a Run 1 
Chamita SNOTEL CSIRO-Mk3-6-0 Run 6 CSIRO-Mk3-6-0 Run 7 MRI-CGCM2.3.2aRun 5 BCC-CSM-1-1 
Cumbres Trestle SNOTEL CSIRO-Mk3-6-0 Run 6 MIROC-ESM-CHEM MRI-CGCM2.3.2aRun 5 CGCm3.1a Run 1 
Big Horn RAWS IPSL-CM5A-LR Run 3 GFDL-CM3 MRI-CGCM2.3.2aRun 5 BCC-CSM-1-1 
Lily Pond SNOTEL CSIRO-Mk3-6-0 Run 6 CSIRO-Mk3-6-0 Run 5 MRI-CGCM2.3.2aRun 5 CGCm3.1a Run 1 
Weather station HD HW WW WD 
Del Norte COOP CSIRO-Mk3-6-0 Run 6 CanESM2 Run 5 MRI-CGCM2.3.2aRun 5 GFDL-ESM2M 
Wolf Creek SNOTEL IPSL-CM5A-LR Run 2 CSIRO-Mk3-6-0 Run 5 MRI-CGCM2.3.2aRun 5 CGCm3.1a Run 1 
Chamita SNOTEL CSIRO-Mk3-6-0 Run 6 CSIRO-Mk3-6-0 Run 7 MRI-CGCM2.3.2aRun 5 BCC-CSM-1-1 
Cumbres Trestle SNOTEL CSIRO-Mk3-6-0 Run 6 MIROC-ESM-CHEM MRI-CGCM2.3.2aRun 5 CGCm3.1a Run 1 
Big Horn RAWS IPSL-CM5A-LR Run 3 GFDL-CM3 MRI-CGCM2.3.2aRun 5 BCC-CSM-1-1 
Lily Pond SNOTEL CSIRO-Mk3-6-0 Run 6 CSIRO-Mk3-6-0 Run 5 MRI-CGCM2.3.2aRun 5 CGCm3.1a Run 1 

aIndicates CMIP phase 3 model. Otherwise models are from CMIP phase 5. Bold entries indicate that the same downscaled model was statistically selected for more than one weather station to represent a specific future state.

CanESM2; Canadian Centre for Climate Modeling and Analysis; CGCm3.1 Canadian Centre for Climate Modeling and Analysis; CSIRO-Mk3-6-0 Commonwealth Scientific and Industrial Research Organization, Queensland Climate Change Centre of Excellence; GFDL-CM3 NOAA Geophysical Fluid Dynamics Laboratory; GFDL-ESM2M NOAA Geophysical Fluid Dynamics Laboratory; IPSL-CM5A-LR Institut Pierre-Simon Laplace; MIROC-ESM-CHEM Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research (The University of Tokyo), and National Institute for Environmental Studies; MRI-CGCM2.3.2 Meteorological Research Institute, Japan.

Table 4

Predicted changes in temperature and precipitation for (1990–1999) and (2090–2099) for five basins of the URG

  HD HW WW WD 
Basin Temperature °C Precipitation % Temperature °C Precipitation % Temperature °C Precipitation % Temperature °C Precipitation % 
Del Norte 5.4 –15.7 6.1 20.7 3.2 16.9 4.4 –14.6 
Chama 6.3 –15.9 6.0 17.8 3.0 14.1 4.2 –17.7 
Conejos 6.1 –15.4 6.2 26.5 3.1 25.7 4.5 –13.5 
Alamosa 6.3 –15.9 6.0 22.6 3.1 17.2 4.3 –13.2 
Los Pinos 6.2 –18.2 6.3 23.7 3.1 23.5 4.5 –17.3 
  HD HW WW WD 
Basin Temperature °C Precipitation % Temperature °C Precipitation % Temperature °C Precipitation % Temperature °C Precipitation % 
Del Norte 5.4 –15.7 6.1 20.7 3.2 16.9 4.4 –14.6 
Chama 6.3 –15.9 6.0 17.8 3.0 14.1 4.2 –17.7 
Conejos 6.1 –15.4 6.2 26.5 3.1 25.7 4.5 –13.5 
Alamosa 6.3 –15.9 6.0 22.6 3.1 17.2 4.3 –13.2 
Los Pinos 6.2 –18.2 6.3 23.7 3.1 23.5 4.5 –17.3 
Figure 2

Mean percent temperature and precipitation difference (1990–1999) vs. (2090–2099) for CMIP3-A2 and CMIP5-RCP8.5 models at six weather stations of the URG Basin.

Figure 2

Mean percent temperature and precipitation difference (1990–1999) vs. (2090–2099) for CMIP3-A2 and CMIP5-RCP8.5 models at six weather stations of the URG Basin.

Effects of climate change on snow cover

The April to September average SCA (%) for the lowest elevation zone (1) ranged from 1 to 12% prior to climate change. The intermediate elevation zone (2) had an average snow cover ranging from 18 to 38%. The highest elevation zone also had the highest average April to September SCA (24–50%). These values are averaged for the three simulation years and presented in Table 5.

Table 5

Average April–September SCA (%) computed using Landsat TM and projected as a result of changing temperatures for four future scenarios

Elevation zone (m) CDC (%) HW (%) HD (%) WW (%) WD (%) 
Alamosa 
 Zone 1 
 Zone 2 15 
 Zone 3 33 17 10 
Chama 
 Zone 1 
 Zone 2 29 15 
 Zone 3 43 13 10 25 16 
Conejos 
 Zone 1 
 Zone 2 19 
 Zone 3 41 19 18 28 22 
Los Pinos 
 Zone 1 
 Zone 2 18 
 Zone 3 40 18 18 27 21 
Del Norte 
 Zone 1 
 Zone 2 18 
 Zone 3 33 15 
Elevation zone (m) CDC (%) HW (%) HD (%) WW (%) WD (%) 
Alamosa 
 Zone 1 
 Zone 2 15 
 Zone 3 33 17 10 
Chama 
 Zone 1 
 Zone 2 29 15 
 Zone 3 43 13 10 25 16 
Conejos 
 Zone 1 
 Zone 2 19 
 Zone 3 41 19 18 28 22 
Los Pinos 
 Zone 1 
 Zone 2 18 
 Zone 3 40 18 18 27 21 
Del Norte 
 Zone 1 
 Zone 2 18 
 Zone 3 33 15 

The largest decline in SCA occurs in the highest elevation zone in each subbasin. Hotter simulations consistently had a larger decline in SCA than warmer simulations. SCA in the highest zone for the Del Norte was eliminated during future climate simulations except the WW simulation which retained 11% SCA.

Effects of climate change on runoff volume and streamflow

Total annual volume results for three years (1996–1999) were averaged to yield the percent difference in volume for each climate scenario and basin. Despite an increase in precipitation of up to 27% in the wetter scenarios, future annual volume only increases by up to 7% (Figure 3). In four of the ten ‘wetter’ simulations, the percent difference from base simulation shows no change or a decrease in total annual volume of −1 to −9%. When drier conditions are simulated, future runoff volume decreases by −17 to −31%. Simulating the bounds of climate model uncertainty provides four realistic, plausible future conditions. The small increase associated with increased precipitation in some years is outweighed by the impacts of increased temperature on available water and results provide a range of conditions for future planning. Chama and Del Norte basins were slightly less impacted by future conditions than Los Pinos and Alamosa, both of which showed a decrease in volume in the HW future condition.
Figure 3

Percent difference from base simulation of total annual runoff volume (106 × m3) averaged over three years (1996–1999) for five URG basins simulated with HD, WD, HW and WW future climate scenarios.

Figure 3

Percent difference from base simulation of total annual runoff volume (106 × m3) averaged over three years (1996–1999) for five URG basins simulated with HD, WD, HW and WW future climate scenarios.

Analysis of monthly runoff volume for each of the scenarios, basins and years provides insight into commonalities and shifts in runoff timing related to increased temperature and altered precipitation (Figure 6). The mean percent increase in runoff volume in March across all basins and years (n = 15) is 246% (WW; standard deviation (SD) = 177%) and 317% (HD; SD = 299%) for wetter scenarios and 193% (WD; SD = 177%) and 168% (HD; SD = 192%) for drier scenarios. April total volume is larger in a changed climate for all models, years and scenarios. Mean percent increase in April runoff volume ranges from 142% (WW; SD = 48%) to 244% (HW; SD = 103%). In May, runoff volume increases for some basins and years and decreases for others. The mean percent change for all basins and years is 3% more for the WW simulation to 30% less for the HD scenario. By June all scenarios and years indicate a decrease in monthly runoff volume with the mean decline (n = 15) ranging from –56% (WW; SD = 11%) to –80% (HD; SD = 9%). Similarly, July volume decreases for all basins, scenarios and years with the mean percent decrease ranging from –38% (WW) to –61% (HD). Unlike prior months, percent change in August and September volume coincides with simulation type with most ‘wetter’ simulations showing an increase in monthly volume and all ‘drier’ simulations showing a decrease in August and September volume. This is likely attributable to the lack of snowmelt runoff in August and September influencing total monthly volume. The simulated increase or decrease in precipitation directly influences monthly volume in August and September.

Effects of shifting the snow runoff coefficient

In this study we shift one parameter shown previously to have a large impact on runoff (Rango & van Katwijk 1991). Daily hydrographs show the earlier runoff and the impact of the cS shift on streamflow (Figures 4 and 5). Review of daily hydrographs for the simulation period reveals both the impact of climate change on streamflow and the effects of shifting cS. Simulating climate change with no parameter shift causes earlier and diminished peak flows at Alamosa Basin for the HD scenario. Peaks are further diminished by shifting cS.
Figure 4

Streamflow (m3/s) at Rio Chama with and without shifting cs for WW scenario.

Figure 4

Streamflow (m3/s) at Rio Chama with and without shifting cs for WW scenario.

Figure 5

Streamflow (m3/s) at Rio Chama with and without shifting cs for HD scenario.

Figure 5

Streamflow (m3/s) at Rio Chama with and without shifting cs for HD scenario.

Three of the five basins (Alamosa, Los Pinos and Conejos) produced similar results, in that shifting cS resulted in an increase in total runoff volume for the HD scenario and a decrease in total runoff volume for the WW scenario. The total runoff volume over three simulation years increased by 2.8 × 106 m3 (HD) and decreased by 3.6 × 106 m3 (WW) for the Alamosa Basin. This represents only a small change in total volume. In contrast, shifting cS for Los Pinos Basin resulted in an increase in simulated total volume of 18.6 × 106 m3 (HD) and this represents 9% of the total runoff, likely an impactful volume when planning for hotter and drier future conditions. Shifting the snow runoff coefficient by 30 days for the Conejos Basin caused a 1.4% increase in total annual volume (HD) to a 2% decrease in total simulated volume (WW). For Alamosa, Los Pinos and Conejos, adjusting cS makes volume for the WW simulation, generally the scenario with the largest simulated volume, lower and the HD simulation (generally the lowest simulated volume) higher, thereby dampening the previously simulated impact of climate change. Failure to include such shifts in future simulations will likely lead to overestimation of total annual volume during a predicted warmer and wetter year and underestimation in a predicted hotter and drier future year.

Shifting cS reduces simulated total volume for each future condition (WW, WD, HW and HD) for Rio Chama. Failure to shift cS would overestimate volume by an annual mean of 3 × 106 m3 in HD conditions and 22 × 106 m3 in WW conditions. Here, shifting cS results in a 1–7% decline in total volume. In contrast to Rio Chama, shifting cS increased simulated three year total volume by 44 (HD), 45 (WD), 48 (HW) and 53 (WW) × 106 m3 or approximately 15–18 × 106 m3 per year. While this is a small percentage of the total annual flow, it indicates that shifting cS could improve simulations by better representing the physical processes linked with earlier increased temperatures associated with climate change.

The small change in Alamosa total volume as compared with other basins is related to the cS values of the initial simulations. Shifting cS only altered April values for one zone (zone 3) and by only 0.05 (from 0.50 to 0.55). In contrast, shifting Los Pinos cS values changed April values from 0.3 to 0.45 for all zones as reflected in the results. Future applications should be mindful of initial cS values and how those might relate to physical changes of increased temperatures. For example, recent simulations use higher cS in June–August than the remainder of the year. In climate change simulations they neglect to shift cS. Their climate change results for a 4 °C increase in temperature show a very large increase in June, July and August streamflow that would likely be somewhat dampened in the late season with the cS shift (Tahir et al. 2011).

To this point we have focused on the impact of shifting cS on total annual volume in five basins. Figure 6 depicts the impacts on monthly runoff volume of the HD and WW scenarios. In general, shifting cS causes the months with higher volume in a changed climate (April and sometimes May) to have higher volume as compared with the base simulation. Shifting cS also leads to months with lower volume in a changed climate (sometimes May, June and July) to have even lower volume as compared with the base simulation. This is especially true for Rio Chama and Los Pinos subbasins. For the WW scenario shifting cS caused a 30% increase in total volume in April and a 200% decrease in total volume in May and a 15% decrease in total June volume. Los Pinos monthly results follow a similar pattern with a 55% increase in total April volume and a 59% decrease in total May volume.
Figure 6

Monthly changes in total volume (106 m3) from base simulations with and without shifting the runoff coefficient for HD and WW scenarios in the URG Basin.

Figure 6

Monthly changes in total volume (106 m3) from base simulations with and without shifting the runoff coefficient for HD and WW scenarios in the URG Basin.

Hydrologic modeling of the impacts of climate change relies upon data of varying certainty. Here we assume that greenhouse gas concentrations will continue to rise during the 21st century and use a more aggressive future condition than others available. We also assume that by statistically selecting models representing a range of future temperature and precipitation change, we most accurately provide simulated changes in daily streamflow. While there is considerable certainty in future temperature changes, precipitation is more variable and anticipated changes are less certain. SRM simulates only changes in temperature, precipitation and parameter timing, assuming these to be the major factors influencing runoff from snowmelt basins. Thus, results presented here omit other factors related to climate change, such as changes in particulate matter, dust on snow, sublimation, and cloud cover. Some of these factors are included in the GCM and thereby influence the temperature and precipitation changes used here. Future climate simulations assumed in SRM generally apply an average temperature and precipitation change for the entire simulation period, however those factors will likely be more variable on a seasonal or sub-seasonal basis.

CONCLUSIONS

We use a methodology to statistically subsample all GCM to inform hydrologic simulation of climate change with a public domain, widely used hydrologic model. Interest in climate change and SRM accessibility, history and ease-of-use leads to much recent use of SRM to simulate impacts of climate change on hydrology of snow-fed systems. The subsample method presented here allows researchers to conduct a feasible number of simulations.

Average SCA in the changed climate decreased the most in highest elevation zones and for the HD simulation and the least for the WW simulation. Despite an increase in ‘wetter’ scenario precipitation of up to 27%, future annual volume only increased by up to 7%. When drier conditions are simulated, future annual runoff volume decreases by −17 to −31%. Despite the large range of predicted future precipitation, the temperature effect of both ‘warmer’ and ‘hotter’ simulations suggests that even with possible future increases in precipitation, total runoff volume will likely decline, or increase only slightly in wet years.

With this work we clearly show the need to shift parameters in future application of SRM in evaluating climate change. The limitations of this study relate to a general inability of most hydrologic models, including SRM, to take full advantage of the advances in available climate change information. With the many GCM and advances in statistical downscaling, there is more data available to simulate future hydrologic conditions. Here we assume the most aggressive future condition, but should emissions and greenhouse gas concentrations decline, then other models may better describe future temperatures. Limitations in computing resources led to selection of four likely future conditions. The annual time scale of the current SRM limits the ability to best use daily temperature data and improvements are in progress to increase the compatibility between SRM and climate predictions, and simultaneously, advances in modeling using artificial neural networks and alternate data pre-processing techniques which may be applied to future climate change simulations (Wu et al. 2009).

In simulating changes in temperature and precipitation in five adjacent basins we expand the knowledge regarding SRM best practices by shifting the snow runoff coefficient while simulating the bounds of plausible future conditions. While our results are consistent with previous efforts in that climate change increases April runoff volume and decreases June and July runoff volume, we show that parameter shifts do not equally influence all basins and that scenario influences the impact of shifting cS. Three of the five basins (Alamosa, Los Pinos and Conejos) produced similar results in that shifting cS resulted in an increase in total runoff volume for the HD scenario and a decrease in total runoff volume for the WW scenario. In this study, shifting cS causes up to a 9% increase in total runoff volume in HD simulations. Monthly simulations of climate change with shifted cS show higher increases in already increased monthly volume (April) and lower decrease in already low monthly volume (June and July). The snowmelt runoff coefficient represents all losses from snowmelt within the basin, such as the physical processes of evaporation and transpiration. Early warm temperatures cause an early onset of these physical processes which must be accounted for by representing this shift in model parameters. As researchers estimate the impacts of climate change on water resources with SRM in basins around the globe, considering the physical basis of parameter values and employing necessary parameter shifts is advised. Based upon past SRM simulations, there are two feasible broad options regarding this component of model application, shifting one or more parameters to reflect temperature changes or neglecting this shift. SRM allows the user to select the number of days to apply a shift to a specific parameter, so the modeler has many options, all of which could affect simulation results. Future simulations shifting other parameters, especially the degree-day coefficient, may further impact climate change simulations. Given the large proportion of the annual URG volume provided by the five simulated subbasins, the results presented here to estimate the impacts of a range of future climate conditions using statistically downscaled GCM data and SRM best practices will inform future water planning and management.

ACKNOWLEDGEMENTS

Downscaling information is based upon work supported by the USBR WaterSmart program (#R11Ap81455; Mejia) and the Desert Research Institute (DRI). We also acknowledge the modeling groups for making their model output available for analysis: the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for collecting and archiving this data and the WCRP's Working Group on Coupled Modelling (WGCM) for organizing the model data analysis activity and BCCA for their downscaling efforts. Portions of this study were funded by the National Science Foundation grant #814449: New Mexico EPSCoR. We thank Jaroslav Martinec for his help in parameterising the basins simulated here and Ryann Smith for project support.

REFERENCES

REFERENCES
Abudu
S.
Cui
C.
Saydi
M.
King
P.
2012
Applications of snowmelt runoff model (SRM) in mountainous watersheds: a review
.
Water Sci. Eng.
5
(
2
),
123
136
.
Bartolino
J. R.
Cole
J. C.
2002
Ground-water resources of the middle Rio Grande Basin
.
USGS Circular 1222, US Geological Survey
,
Reston, VA
.
Boudhar
A.
Hanich
L.
Boulet
G.
Duchemin
B.
Berjamy
B.
Chehbouni
A.
2010
Evaluation of the snowmelt runoff model in the Moroccan High Atlas Mountains using two snow-cover estimates
.
Hydrol. Sci. J.
54
,
1
20
.
Cayan
D.
Tyree
M.
Kunkel
K.
Castro
C.
Gershunov
A.
Barsugli
J.
Ray
A. J.
Overpeck
J.
Anderson
M.
Russell
J.
Rajagopalan
B.
Rangwala
I.
Duffy
P.
2013
The southwest climate of the future – projections of mean climate
. In:
Assessment of Climate Change in the Southwest United States: A Report Prepared for the National Climate Assessment
(
Garfin
G.
Jardine
A.
Merideth
R.
Black
M.
LeRoy
S.
, eds).
NCA Regional Input Reports
.
Island Press
,
Washington, DC
.
Colorado Division of Water Resources
2008
.
DeWalle
D. R.
Rango
A.
2008
Principles of Snow Hydrology
.
Cambridge University Press
.
Cambridge
,
UK
, p.
410
.
Elias
E. H.
Rango
A.
Steele
C. M.
Mejia
J. F.
Smith
R.
2015
Assessing climate change impacts on water availability of snowmelt-dominated basins of the Upper Rio Grande basin
.
J. Hydrol.
3
,
525
546
.
Environmental Systems Resource Institute
2010
ArcGIS. ESRI, Redlands, California. Available from: www.esri.com
.
Garfin
G.
Jardine
A.
Merideth
R.
Black
M.
LeRoy
S.
(eds)
2013
Assessment of Climate Change in the Southwest United States: A Report Prepared for the National Climate Assessment
.
A report by the Southwest Climate Alliance
.
Island Press
,
Washington, DC
, p.
506
.
Gesch
D. B.
2007
The National Elevation Dataset
.
In:
Digital Elevation Model Technologies and Applications: The DEM Users Manual
(
Maune
D.
, ed.).
American Society for Photogrammetry and Remote Sensing, Bethesda
,
Maryland
, pp.
99
118
.
Gesch
D.
Oimoen
M.
Greenlee
S.
Nelson
C.
Steuck
M.
Tyler
D.
2002
The National Elevation Dataset
.
Photogramm. Eng. Remote Sens.
68
(
1
),
5
11
.
Gudmundsson
L.
Bremnes
J. B.
Haugen
J. E.
Engen-Skaugen
T.
2012
Technical note: downscaling RCM precipitation to the station scale using statistical transformations–a comparison of methods
.
Hydrol. Earth Syst. Sci.
16
(
9
),
3383
3390
.
Gutzler
D. S.
Garfin
G.
Zak
B.
2006
Observed and predicted impacts of climate change on New Mexico's water supplies
. In:
The Impact of Climate Change on New Mexico's Water Supply and Ability to Manage Water Resources
(
Watkins
A.
, ed.).
New Mexico Office of the State Engineer/Interstate Stream Commission
,
Santa Fe, New Mexico
, pp.
4
32
.
Haerter
J. O.
Eggert
B.
Moseley
C.
Piani
C.
Berg
P.
2015
Statistical precipitation bias correction of gridded model data using point measurements
.
Geophys. Res. Lett.
42
(
6
),
1919
1929
. doi: 10.1002/2015gl063188.
Hall
D. K.
Riggs
G. A.
Salomonson
V. V.
2006
MODIS/Terra Snow Cover Daily L3 Global 500 m Grid, Version 5. Available from: http://nsidc.org/data/mod10a1
.
Knutti
R.
Abramowitz
G.
Collins
M.
Eyring
V.
Gleckler
P. J.
Hewitson
B.
Mearns
L.
2010
Good practice guidance paper on assessing and combining multi model climate projections
. In:
Meeting Report of the Intergovernmental Panel on Climate Change Expert Meeting on Assessing and Combining Multi Model Climate Projections
(
Stocker
T. F.
Qin
D.
Plattner
G.-K.
Tignor
M.
Midgley
P. M.
, eds).
IPCC Working Group I Technical Support Unit, University of Bern
,
Bern
,
Switzerland
.
Kundzewicz
Z. W.
Mata
L. J.
Arnell
N. W.
Döli
P.
Jimenez
B.
Miller
K.
Oki
T.
Şen
Z.
Shiklomanov
I.
2008
The implications of projected climate change for freshwater resources and their management
.
Hydrol. Sci. J.
53
,
3
10
.
Llewellyn
D.
Vaddey
S.
2013
West-Wide Climate Risk Assessment: Upper Rio Grande Impact Assessment
.
Albuquerque
,
New Mexico
. .
Martinec
J.
1975
Snowmelt-runoff model for streamflow forecasts
.
Nord. Hydrol.
6
(
3
),
145
154
.
Martinec
J.
1980
Snowmelt Runoff Forecasts Based on Automatic Temperature Measurements
.
International Association of Hydrological Sciences Symposium
,
Oxford
,
IAHS-AIHS Publ. no. 129. Accessed online: http://iahs.info/uploads/dms/iahs_129_0239.pdf
.
Martinec
J.
Rango
A.
Roberts
R.
2008
Snowmelt Runoff Model (SRM) User’s Manual
.
New Mexico State University
,
Las Cruces, NM
. .
Maurer
E. P.
Hidalgo
H. G.
Das
T.
Dettinger
M. D.
Cayan
D. R.
2010
The utility of daily large-scale climate data in the assessment of climate change impacts on daily streamflow in California
.
Hydrol. Earth Syst. Sci.
14
,
1125
1138
.
McCuen
R.
Knight
Z.
Cutter
A.
2006
Evaluation of the Nash–Sutcliffe efficiency index
.
J. Hydrol. Eng.
11
(
6
),
597
602
.
McEvoy
D. J.
Mejia
J. F.
Huntington
J. L.
2014
Use of an observation network in the great basin to evaluate gridded climate data
.
J. Hydrometeor.
15
,
1913
1931
.
Mejia
J. F.
Huntington
J.
Hatchett
B.
Koračin
D.
2011
Linking global climate models to an integrated hydrologic model using a hybrid downscaling approach
.
J. Contemp. Water Res. & Educ.
147
,
17
27
.
Mejia
J. F.
Huntington
J.
Niswonger
R.
2014
Uncertainty transfer in modeling layers: from GCM to downscaling to hydrologic surface-groundwater modeling
. In:
Proceedings of the 7th International Congress on Environmental Modelling and Software, June 15–19
(
Ames
D. P.
Quinn
N. W. T.
Rizzoli
A. E.
, eds).
San Diego, California
,
USA
. .
Moriasi
D. N.
Arnold
J. G.
Van Liew
M. W.
Harned
R. D.
2007
Model evaluation guidelines for systematic quantification of accuracy in watershed simulations
.
Trans. Am. Soc. Agric. Biol. Eng.
50
(
3
),
885
900
.
Nash
L. L.
Gleick
P. H.
1993
The Colorado River Basin and Climatic Change: the Sensitivity of Streamflow and Water Supply to Variations in Temperature and Precipitation
.
US Environmental Protection Agency EPA230-R-93-009
.
Pacific Institute for Studies in Development, Environment and Security, Oakland, CA
.
National Oceanic and Atmospheric Administration
2013
National Climatic Data Center, Cooperative Observer Network (COOP) Data. Available from: www.ncdc.noaa.gov/data-access
.
Overpeck
J.
Garfin
G.
Jardine
A.
Busch
D. E.
Cayan
D.
Dettinger
M.
Fleishman
E.
Gershunov
A.
MacDonald
G.
Redmond
K. T.
Travis
W. R.
Udall
B.
2013
Summary for decision makers
. In:
Assessment of Climate Change in the Southwest United States: A Report Prepared for the National Climate Assessment
(
Garfin
G.
Jardine
A.
Merideth
R.
Black
M.
LeRoy
S.
, eds).
A report by the Southwest Climate Alliance
.
Island Press
,
Washington, DC
. pp.
1
20
.
Perkins
S. E.
Irving
D. B.
Brown
J. R.
Power
S. B.
Moise
A. F.
Colman
R. A.
Smith
I.
2012
CMIP3 Ensemble climate projections over the western tropical Pacific based on model skill
.
Clim Res.
51
,
35
58
.
Rango
A.
2006
Snow: The real water supply for the Rio Grande Basin
.
N. M. J. Sci
.
44
,
99
118
.
Rango
A.
Martinec
J.
1981
Accuracy of snowmelt runoff simulation
.
Nord. Hydrol.
12
,
265
274
.
Rango
A.
Martinec
J.
1994
Areal extent of seasonal snow cover in a changed climate
.
Nord. Hydrol.
25
,
233
246
.
Rango
A.
Martinec
J.
1997
Water storage in mountain basins from satellite snow cover monitoring
. In:
Remote sensing and geographic information systems for design and operation of water resources systems, Rabat, Morocco. International Association of Hydrological Sciences
, pp.
83
91
.
Rango
A.
Martinec
J.
1999
Modeling snow cover and runoff response to global warming for varying hydrological years
.
World Resour. Rev.
11
(
1
),
76
91
.
Rango
A.
Martinec
J.
2000
Hydrological effects of a changed climate in humid and arid mountain regions
.
World Resour. Rev. (Glob. Warm. Int. Ctr.M.)
12
(
3
),
493
508
.
Rango
A.
van Katwijk
V. F.
1990
Climate change effects on the snowmelt hydrology of Western North American Mountain Basins
.
IEEE Trans. Geosci. Remote Sens.
28
(
5
),
970
974
.
Rango
A.
van Katwijk
V. F.
1991
Water Supply Implications of Climate Change in Western North American Basins. International and Transboundary Water Resources Issues
.
American Water Resources Association
,
Middleburg, VA
, pp.
557
586
.
Reclamation
2008
Sensitivity of Future CVP/SWP Operations to Potential Climate Change and Associated Sea Level Rise, Appendix R in Biological Assessment on the Continued Long-term Operations of the Central Valley Project and the State Water Project. Prepared by Bureau of Reclamation, US Department of the Interior, August 2008
, p.
134
.
Reclamation
2013
Downscaled CMIP3 and CMIP5 Climate and Hydrology Projections: Release of Downscaled CMIP5 Climate Projections, Comparison with Preceding Information, and Summary of User Needs
.
Prepared by the US Department of the Interior, Bureau of Reclamation
,
Technical Services Center
,
Denver, Colorado
, p.
47
.
Seidel
K.
Martinec
J.
Baumgartner
M. F.
2000
Modeling runoff and impact of climate change in large Himalayan basins
In:
Integrated Water Resources Management for Sustainable Development, Vol. 2
.
National Institute of Hydrology
,
Roorkee, New Delhi
,
India
, pp.
1020
1028
.
Sheffield
J.
Camargo
S. J.
Fu
R.
Hu
Q.
Jiang
X.
Johnson
N.
Karnauskas
K. B.
Kim
S. T.
Kinter
J.
Kumar
S.
Langenbrunner
B.
Maloney
E.
Mariotti
A.
Meyerson
J. E.
Neelin
J. D.
Nigam
S.
Pan
Z.
Ruiz-Barradas
A.
Seager
R.
Serra
Y. L.
De-Zheng Sun
D. Z.
Wang
C.
Xie
S. P.
Yu
J. Y.
Zhang
T.
Zhao
M.
2013
North American climate in CMIP5 experiments. Part II: evaluation of historical simulations of intraseasonal to decadal variability
.
J. Clim.
26
,
9247
9290
.
Steele
C. M.
Rango
A.
Elias
E. H.
2015
Evaluating MODIS snow products for modelling snowmelt runoff: case study of the Rio Grande headwaters
.
Int. J. Appl. Earth Observ.
Under Review
.
Stott
P. A.
Forest
C. E.
2007
Ensemble climate predictions using climate models and observational constraints
.
Phil. Trans. R. Soc. A.
365
,
2029
2052
.
US Department of Agriculture, Natural Resources Conservation Service
2013
SNOTEL Data and Products. Available from: www.wcc.nrcS.usda.gov/snow/
.
US Forest Service
2011
Remote Automated Weather Stations Data. Available from: www.raws.dri.edu/
.
US Geological Survey
2001
National Water Information System Data. Available from: http://waterdata.usgs.gov/nwis/
.
US Geological Survey
2012
Shuttle Radar Topography Mission (SRTM) Data. Available from: https://lta.cr.usgs.gov/SRTM2
.
van Katwijk
V. F.
Rango
A.
1991
The effect of climate change on snowmelt runoff in two different climatic regions of the western United States
. In:
Tenth Conference on Biometeorology and Aerobiology and the Special Session on Hydrometeorology
.
September 10–13, 1991
.
Salt Lake City, UT. 210-213
.
American Meteorology Society
,
Boston, MA
.
van Katwijk
V. F.
Rango
A.
Childress
A. E.
1993
Effects of simulated climate change on snowmelt runoff modeling in selected basins
.
Water Resour. Bull.
29
(
5
),
755
766
.
Weigel
A. P.
Knutti
R.
Liniger
M. A.
Appenzeller
C.
2010
Risks of model weighting in multi model climate projections
.
J. Clim.
27
,
4175
4191
.
WMO (World Meteorological Organization)
1986
Intercomparison of Models of Snowmelt Runoff
.
Operational Hydrology. Report 23
.
WMO
,
Geneva
,
Switzerland
.
WMO (World Meteorological Organization)
1992
Simulated Real-Time Intercomparison of Hydrological Models
.
Operational Hydrology. Report 38
,
WMO
,
Geneva
,
Switzerland
.